ProDeo (A Crazy Idea)

[mar]

Fabien Letouzey wrote Fruit in the beginning to mid-2000's. That program reached an Elo-rating of 2695. 10 years later he released Senpai, and that program reached a rating of 3100. That's a 400 point advantage for a newer engine.

Senpai 1.0 is 2961, Senpai 2.0 is 3032 single threaded (CCRL 40/15), 3100 is SMP with 4 cores, do the maths yourself.
Well not everyone is Fabien plus I don't remember him ever claim that the improvements come from upgrading to a newer version of the compiler or similar nonsense.

The point is that to gain elo, you need to make real progress and that requires real effort, obviously.
A rewrite won't get you that per se, it only gives you a better codebase, you then have to build on top of that and add elo the standard way (unless the rewrite is like Python => C).

[AdminX]

But i HAVE mchess, cstal, rebel, isichess, genius, Fritz, ... wchess, ... running in android using magic dos box.


The reason eds program is 2600-2700 while others begin with 2300 and very fast reach 3500 is that they copy ideas from Stockfish.

When ed began programming the sources were all secret.

The problem i see is that each programmer is in a kind of paradigm.
And its difficult to go OUT of this paradigm.

Haha now so do I. Thanks for the tip about Magic Dosbox you gave me years ago. Back then it was all about my phone, but now that I own a Chrome OS tablet, that software has really come in handy. I am missing CSTal , but have Zarkov, and NOW it's like hanging out with old friends!

[Rebel]

Actually I made a (small) start with this 4-5 years ago just for the fun of it, not sure if I still have it, but it's a lot of work and I gave up. OTOH, there will be a new ProDeo soon.

If I understand correctly, ProDeo is actually "Rebel 13", but freeware. Is it also based on the Rebel 12 source code?

Yes.

I retired somewhere in 2003/4 and Rebel (paid) became ProDeo (for free in Dutch), I got my hobby back.

Seeing that ProDeo is rated "only" 2722, have you never thought about just starting a new engine from scratch? I'm sure that if you did, you could easily surpass ProDeo with regard to strength. Of course it would be possible to start replacing small parts of ProDeo with newly written code, and so "phase out" the existing engine without making a hard break. I could imagine though, that Rebel's/ProDeo's source code would make this more work than just starting over.

I've been thinking about this with regard to authors that have been writing chess programs for a long time; is it still possible somehow to improve older engines to compete with newer ones without rewriting half the engine?

To improve playing strength there is no need to write the engine from scratch, just plunder the Stockfish search Instead I prefer to keep a low profile and concentrate on experiments and adding features.

[mvanthoor]

And its difficult to go OUT of this paradigm.

That is one of the things I meant. Older engines are based on older ideas. If an engine is (for exmample) an array-based engine, it's very hard to change it to become a magic bitboard engine. You'd have to rewrite so much code that it's probably better to start over.

Senpai 1.0 is 2961, Senpai 2.0 is 3032 single threaded (CCRL 40/15), 3100 is SMP with 4 cores, do the maths yourself.
Well not everyone is Fabien plus I don't remember him ever claim that the improvements come from upgrading to a newer version of the compiler or similar nonsense.

Of course switching compilers doesn't gain you a few hundred Elo. I know that. But you also state:

The point is that to gain elo, you need to make real progress and that requires real effort, obviously.
A rewrite won't get you that per se, it only gives you a better codebase, you then have to build on top of that and add elo the standard way (unless the rewrite is like Python => C).

A rewrite gives you a better code base. There may be new techniques that Ed CAN'T build into ProDeo, even if he wanted to. (Assuming ProDeo actualy descents directly from Rebel.)

It's very possible that ProDeo (or other engines based on older code) contain code that prevents the author from upgrading compilers, and thus (for example) prevents them from using things like intrinsics, or make it very hard / time consuming / impossible to add newer features due to restrictions on the older version of the language. That sort of code may not be simply removable because it can be an integral part of the engine. It can prevent the engine from being usable on Android for example, because there's no compiler that supports the older code AND the newer Android versions.

I've seen many software projects in embedded engineering that are stuck this way: impossible / hard to to add features, impossible / hard to remove code, impossible to compile for a newer platform. Stuck where they are because of technical debt that can't be resolved any more. The only way to actually make progress is to rewrite the project.

It happens. It would not be a shame for someone like Ed to come to that conclusion, stop development on the current engine (but keep it running on newer platforms where possible), and start something new to see where he can take it. I would applaud it: a new engine from scratch from one of the old masters.

[Madeleine Birchfield]

To improve playing strength there is no need to write the engine from scratch, just plunder the Stockfish search

No, you keep the original engine's search, and replace the evaluation function with a neural network.

[mvanthoor]

I retired somewhere in 2003/4 and Rebel (paid) became ProDeo (for free in Dutch), I got my hobby back.

Cool Sometimes, making work out of your hobby indeed costs you your hobby. (Been there, done that... with photography in my case.)

To improve playing strength there is no need to write the engine from scratch, just plunder the Stockfish search Instead I prefer to keep a low profile and concentrate on experiments and adding features.

What features does the current version of ProDeo have that the Rebel engine didn't have? (Apart from additions to the search and evaluation, I mean.)

[AdminX]

But i HAVE mchess, cstal, rebel, isichess, genius, Fritz, ... wchess, ... running in android using magic dos box.


The reason eds program is 2600-2700 while others begin with 2300 and very fast reach 3500 is that they copy ideas from Stockfish.

When ed began programming the sources were all secret.

The problem i see is that each programmer is in a kind of paradigm.
And its difficult to go OUT of this paradigm.

Haha now so do I. Thanks for the tip about Magic Dosbox you gave me years ago. Back then it was all about my phone, but now that I own a Chrome OS tablet, that software has really come in handy. I am missing CSTal , but have Zarkov, and NOW it's like hanging out with old friends!

Stop the press, I located it (CSTAL).

[towforce]

I've been playing around with some of the old Dos Rebels on my Tablet using Magic Dosbox and started wondering how difficult would it be to backport Prodeo (Maybe even Benjamin) to the old dos Rebel GUI? I know it is a crazy idea, but it would allow users a way to use them under Android via DosEMU.

Have you tried Wine for Android? That might allow ProDeo to run under Android?

[AdminX]

I've been playing around with some of the old Dos Rebels on my Tablet using Magic Dosbox and started wondering how difficult would it be to backport Prodeo (Maybe even Benjamin) to the old dos Rebel GUI? I know it is a crazy idea, but it would allow users a way to use them under Android via DosEMU.

Have you tried Wine for Android? That might allow ProDeo to run under Android?

No, I been using and old network program by Bernhard Wallner that I think it was called NetChess or something like that. It allows me to run a Chess Engine on PC and use Chess For Android on my Android device for Engine vs Engine Matches by using an assigned port to listen for communication between engines. I will check out Wine for Android when I have time as you can never have enough tools in the toolbox. Thanks

[mclane]

I retired somewhere in 2003/4 and Rebel (paid) became ProDeo (for free in Dutch), I got my hobby back.

To improve playing strength there is no need to write the engine from scratch, just plunder the Stockfish search Instead I prefer to keep a low profile and concentrate on experiments and adding features.

I do think it could be possible to increase strength of prodeo.

I remember that Ed tried out zillions of switches and ideas.
And imo one major problem was that with any new version, the switches also changed.
And so the style that „worked“ with Version x suddenly produced negative effects in Version x+1.
Or that certain features that worked in x suddenly had no effect because internal the setting in the style were set off .

The whole project has grown to such a complexity, it would need a genius to understand all those switches and functions.

It might have been easier in times were there was „only“

Mephisto Rebel
Mephisto MM4
Mephisto Monte Carlo
Mephisto Mega 4
Mephisto academy
Mephisto polgar
Mephisto Milano
Tasc Madrid 3.1 / Mephisto Risc 1+2
Mephisto Nigel Short


And not to forget all the experimental versions that existed in between those commercial versions.

But with the rise of pc chess eds engines increased complexity.

That’s my subjective opinion of course and maybe I am wrong.

Maybe beginning from the scratch with a less complex version could really be a solution.